【作者】 周银兴；

【导师】 薛兵；

【作者基本信息】 中国地震局地震预测研究所 ， 固体地球物理， 2009， 硕士

【摘要】 近30年我国数字地震观测台网迅速发展,特别是随着“十五”项目“中国数字地震观测网络项目”的完成,以及近年来经济发展迅猛,国家对于大型可能诱发水库地震的水库、受地震后可能引发次生灾害的油田、矿山、石油化工及大型煤矿等重要设施要求必须建设专用地震监测台网,地震台网规模越来越大,地震观测台站越来越多,数据每年以海量产出,人们对实时地震数据自动处理系统的要求越来越迫切,同时由于地震是自然灾害中破坏性最强的,近年来人们对于建立地震预警系统的呼声也越来越高,这些都是建立在深入研究地震事件和地震震相的自动识别基础之上。微小地震虽然破坏性不大,但却是研究大地震和整个地震带应力场的基础,本世纪初在经历了印尼大地震引发的海啸和“5.12”汶川大地震以后,我们清楚认识到当前对于地震预报这一世界性科学难题仍然没有更好的办法,但对于减轻地震灾害的渴望越来越大,因此国家努力加大台网观测密度,丰富了观测资料,提供了我们进一步认识地下精细结构的可能,小地震震相简单,在目前的技术前提下,比较容易实现自动识别,震相的自动识别处理不但能减轻人们处理资料的负担,而且能快速的获得更多的震相识别结果、便于人们了解地震时空发展动态过程具有深远的意义。地震的自动处理过程包括事件自动检测和震相自动识别两部分,目前没有一种方法可以把这两件事情做得很好,基于目前的情况,我们深入的研究了目前常用的SLT/LSA等检测算法。总结了人们对于地震事件的研究和震相识别的各种方法,国内的应用情况,具有征对性的进行了以下几方面的详细研究:1.对区域地震事件检测方法进行了总结,详细研究了STA/LTA这一应用最广泛的检测方法和偏振(极化)检测法,编写程序验证地震波的偏振对于检测微震事件和判断震相的可行性。基于协方差矩阵的特征值法和奇异值分解的方法可较清晰的反映宽频带记录中的小地震事件,而且是基于三分向记录进行的检测方法,实践表明,该方法可增强识别率降低误检率,可用该方法构造特征函数与别的方法综合进行微震事件检测。2.总结了震相自动识别的方法,以Akaike信息准则(AIC)为基础,用基于自回归的AR-AIC、基于样本方差的VAR-AIC、基于高阶统计量的TOC-AIC、和基于记录振幅(能量)的AOC-AIC方法对比进行自动检测了小地震中的PG、SG震相,自动识别的结果表明,较高的信噪比有较高的识别精度,低信噪比事件数据需要经过滤波处理后方能得到可靠的自动识别精度。各种AIC方法中,VAR-AIC、AOC-AIC方法计算量小,速度快,应用AIC方法识别震相到时必须保证计算的数据段便能保证识别的可靠性,AIC方法是基于单分向数据进行检测的。偏振特征法是基于三分向数据进行检测和震相分析的,具有较好的可信度,但是因检测识别计算过程中要取一定长的时间窗口构造协方差矩阵,因此在进行震相识别的时候所取数据长度窗口越小,震相起始点的精度越高。3.我们应用P波与S波的质点运动特征,以乌江水库地震台网的资料分析了P波段、S波段质点偏振主轴方向Z与其它两个正交方向Y与X的特征关系进行了分析统计,结果表明以P波段为偏振模型进行坐标旋转以后与S波三分向特征值特点差异明显,P波质点运动主轴方向Z与另两正交方向上的特征值的幅值比2×Z / ( X + Y)、Y与X的幅值比、Y与主轴方向Z的幅值比均有不同的分布特征,P波与S波主轴方向与两正交方向特征值的比值2×Z / ( X +Y)差异最大,P波主轴方向的方位角入射角信息与S波也有明确的分界,这些偏振特征可作为自动检测结果的判定标准模型。更多还原

【Abstract】 With the rapid development of economic and seismic observation network, especially, after China Earthquake Administration finishes constructing the“China digital earthquake observation network project”of the“five year plan”of Chinese central government. There are about 800 professional seismologic observation and many of local and enterprise seismologic stations have been built up for monitoring the natural and man-made earthquake, because every station is real-time sending and recording the seismic data, there are a lot of record data accumulated. The analyzers always expecting the automatic processing come true and the method is credible. We all know about the huge earthquake happened in Indonesia in 2004 and the“5.12”big earthquake happened in Sichuan in 2008. Big earthquake can destroy everything and the scientist still can not forecast, so people have to pay more attention to build the earthquake warning system. No matter the real-time automatic processing system and early warning system, they are all base on the credible automatic seismic event detection and phase identification.This paper sums up the general seismic event detection and phase identification methods. Focusing on the STA/LTA check arithmetic, earthquake polarization character and base on Akaike information criterion for application on small earthquake detection and PG and SG phase identification. What I have done is:1. Briefly summarized some seismic event detection methods, analyzed the STA/LTA detection method and polarization character. The research results indicates: the eigvalue method which is based on the covariance of the three channel record data can distinguish the noise and seismic clearly, SVD (singularity value decompose) method can also show this character and can show cleaner when the small earthquake recorded by broadband seismometer.2. In this paper, I summarized the phase identification method used by researchers, and continue research the AR-AIC(automatic regressive)、VAR-AIC(based on the variance)、TOC(High)-AIC(tall or high order statistics)、AOC-TOC(based on Amplitude or Power), compared the different AIC method detect result. The VAR-AIC and AOC-AIC method need the least time and calculate fastest. When the SNR is high, every method can get the satisfying PG and SG start time. The polarization method is based on three channel seismic record, the detection result is credible, but when use the method to identify the phase the result is lie on the data window you chosen. the window is small the result is higher.3. We use the P wave data window and S wave data window of three components record get the principal axes of the polarization ellipsoid respectively, Using 175 three channel events records of WuJiang reservoir telemetering seismic network we get the Vertical/Horizontal, (2*Z/(Y+X))、Vertical/Radial(Z/Y)、Radial /Transverse(Y/X) ratio statistics distributing characterics of P wave and S wave on their polarization axes, the results show Z/H ratios has the most difference, the V/R and R/T ratio also have obvious difference, We still take the P wave’s polarization axes as the model standard, rotate the subsequent data of Pg phase, then detect the phase using AIC method automatically. We hope this will help us identify the automatic phase detection result.更多还原